JPH0410109B2 - - Google Patents

Description

[Detailed Description of the Invention] [Summary] In machine translation, in order to select the most suitable translation word according to the meaning of the sentence, the priority of words is dynamically set based on the results of previous translations in the same sentence. This will enable the correct selection of translation words.

[Industrial Application Field] The present invention relates to selection of translated words in machine translation. In machine translation, an important measure that determines the quality of the results is whether the translated words are correct.
Therefore, in order to select the correct translation, if multiple translations correspond to the same word, it is necessary to dynamically change the priority during translation.

[Conventional technology]

A translation device performs sentence analysis such as morphological analysis, syntactic analysis, and semantic analysis to generate a conceptual structure, and generates a sentence based on the conceptual structure.

The purpose of morphological analysis is to find the morphemes that make up an input sentence. Morphological analysis is performed using a word dictionary and a connection table in which the possibility of adjacency between words is entered.

Syntactic analysis determines how the elements of a sentence are connected.
This clarifies what role they play. That is, grammatical rules are applied to the elements of a sentence broken down into morphemes to determine the composition of clauses and phrases, and the modification relationships between subjects and predicates. However, at this stage, even if the analysis is grammatically correct, it is not necessarily semantically correct. Therefore, the following semantic analysis is required.

In semantic analysis, the semantically correct one is selected from the grammatically correct analysis results. In syntactic analysis, there are multiple correct results depending on the ambiguity of the word. This includes some sentences that have completely strange meanings, so if you perform a semantic analysis during the syntactic analysis and filter out the words that have strange meanings, you can perform an efficient analysis.

A conceptual structure is a representation of the meaning of a sentence in the form of a semantic network. This conceptual structure is created so as not to depend on the language to be translated, so it is possible to generate sentences in multiple languages from this conceptual structure.

Sentence generation is divided into syntactic generation and morpheme generation. In the process of generating sentences in the target language (English in the case of Japanese-English translation) from conceptual structures, words are extracted from the dictionary of the target language based on the conceptual symbols that express the concepts. At this time, since grammatical information is not indexed at the conceptual structure level, it is necessary to select the part of speech that matches the target language. These rules are called production rules. In production rules, processing starts from the node where the arc of the entry point shown in the conceptual structure enters, and the syntax of that node is determined by looking at the relationship between the surrounding arcs, the nodes ahead of it, and the grammatical attributes of the selected word. Co-occurrence relationships are used to determine phrasing between words for the syntax determined by syntax generation. The co-occurrence relationship determines which word to select when multiple words exist for one concept. In morpheme generation, for words determined through syntactic generation and co-occurrence relationships, the inflectional part of the word, which can be determined based on information such as tense, person, number, case, gender, etc., is generated by referring to a word dictionary. Compose words with meaning.

The present invention relates to a method of selecting words corresponding to conceptual symbols from a word dictionary. FIG. 3 shows a conventional word selection method. In FIG. 3, 1 indicates a word dictionary, and 2 indicates a dictionary access device. In the word dictionary 1, for example, corresponding to the conceptual symbol "line", line, track,
English words such as route, wire, policy, string, etc. and the priority of these English words are entered. In the illustrated example, the priority of line is 100, and the priority of track is 50, for example. The dictionary access device 2 is for accessing the word dictionary 1. The priority corresponding to each word is fixed. Upon receiving the instruction to search for an English word corresponding to “line,” the dictionary access device 2 searches the word dictionary 1 for line, track, route, wire, etc.
Reads English words such as policy and string, arranges them in descending order of priority, and passes them to the requester.

[Problem to be solved]

With conventional technology, when multiple translations correspond to one conceptual symbol, the priority of the translations does not change dynamically according to the input text, so the correct translation is not given in the translation result. It had to be corrected by a human.

The present invention was created in view of this point, and an object of the present invention is to enable dynamically changing the priority of translated words.

[Means for solving problems]

FIG. 1 is a diagram showing the principle of the present invention. In FIG. 1, 3 is a translation device, 4 is a priority change device, 5 is a thesaurus dictionary, 6 is an excluded term dictionary, 7 is a translation analysis device, and 8 and 9 are other various dictionaries, respectively.

The translation device 3 performs morphological analysis, syntactic analysis, generation of a semantic network, generation of syntax, generation of morphemes, and the like. Translated words used in the translation process (including when the user intervenes and selects one of a plurality of translated words) are sent to the priority change device 4. When the priority changing device 4 receives the translated word, it sends it to the translated word analyzing device 7. The translation analysis device 7 analyzes the received sentence, decomposes it into words, and returns the words as the analysis result to the priority change device 4. The priority change device 4 refers to the excluded word dictionary 6 to check whether the word sent from the translated word analysis device 7 is an excluded word. The excluded word means a word that is not subject to priority change. If the word is not an excluded word, the priority change device 4
refers to the thesaurus dictionary and determines the increase in priority of the word's broader terms, synonyms, related words, and narrower terms. Information about the increase in priority is stored in the buffer storage area. This storage area can be referenced by the dictionary access device 2. When it is desired to read a word corresponding to a conceptual symbol, the translation device 3 sends a word search request to the dictionary access device 2. When the dictionary access device 2 receives a word search request from the translation device 3, it reads out the word and priority corresponding to the conceptual symbol from the word dictionary 1. Then, the increase in priority corresponding to the word read from the word dictionary 1 is read from the buffer storage area, the priority of the word and the increase in priority of the word are added, and the word is read out according to the updated priority. The words read out from the dictionary 1 are rearranged and these words are passed to the translation device 3. Dictionaries 8 and 9 are other dictionaries. Other dictionaries include adjacency dictionaries, grammatical rules dictionaries,
There are production rule dictionaries, co-occurrence relation dictionaries, semantic relation dictionaries, etc.

〔Example〕

FIG. 2 is a diagram showing an example of changing the priority of words. The input sentence, as shown in (a), is ``Due to the development of airplanes...the structure of the wings...''. The first sentence, “Due to the development of airplanes…” is translated as “the development of…” as shown in (b).
"airplane..." After analyzing the resulting translations, the excluded words are marked with a cross (c). This process is performed by the excluded term dictionary 6 and translated word analysis device 7 of the priority change device 4. The content of the target term airplane in the thesaurus dictionary 5 is (d). is a hypernym of the term airplane
The priority of machine is increased by 10 and also the term
The priority of vehicle, which is a hypernym of airplane, increases by 50, and the priority of aircraft, which is a synonym of airplane, increases by 200 and is a related word of airplane.
The priorities of pilot and airfield will also increase by 200, and the priorities of wing and rudder, which are hyponyms of airplane, will increase by 100. Due to increased priority,
Priority of translated words for feather (feather p (100), plume
(10), wing(5)) changes as shown in (e). The dictionary access device 2 passes this changed list of translated words to the translation device 3. This causes “splash” in the nth sentence.
As shown in (f), the translated word wing is selected for .

〔Effect of the invention〕

As is clear from the above description, according to the present invention, translation words are selected in consideration of the preceding and following sentences, and higher quality translation results can be obtained.

[Brief explanation of drawings]

FIG. 1 is a diagram showing the principle of the present invention, FIG. 2 is a diagram showing an example of changing the priority of words, and FIG. 3 is a diagram showing a conventional word selection method. 1...Word dictionary, 2...Dictionary access device, 3
...Translation device, 4...Priority change device, 5...Thesaurus dictionary, 6...Exclusion word dictionary, 7...Translation analysis device, 8 and 9...Other various dictionaries.

Claims (1)

[Scope of Claims] 1. A translation device that performs machine translation using various dictionaries such as a word dictionary, a grammar rule dictionary, and a thesaurus dictionary, comprising: a translated word analysis device 7 that analyzes a translated sentence and breaks down the order of words; an excluded term dictionary 6 that stores excluded words whose degree is not changed; a thesaurus dictionary 5 that stores hypernyms, synonyms, related words, hyponyms, etc. of words; a priority change device 4; a word dictionary 1; The priority change device 4 requests the translated word analysis device 7 to analyze the input translation result sentence, and when the translated word analysis device 7 returns a word sequence,
With reference to the excluded word dictionary 6, it is checked whether or not the words constituting the word sequence are excluded words, and for words that are not excluded words, the thesaurus dictionary 5 is referred to, and the superordinate words related to these words are checked.
The dictionary access device 2 is configured to obtain the increase in priority for synonyms, related words, and subordinate words, and store the increase in priority and the corresponding word in the increase storage area, and the dictionary access device 2 performs a word search. When a request is received, the word corresponding to the conceptual symbol specified in the word search request and its accompanying priority are read from the word dictionary 1, and the priority corresponding to the read word is stored in the increase storage area. If the increase in priority is stored, the increase in priority is added to the read priority, and the addition result and the read word are passed to the word search request source. A translation device using the context.